This article draws insights from a recent episode of Data Chats featuring Kevin Hanegan, chief learning officer at Qlick >> Listen to the full episode.
When teams start working with data, the earliest mistake they make is failing to identify the right problem or align the problem they are trying to solve with the broader organizational strategy.
A truly effective data project requires more than just technical skills and the right models. With the non-technical elements of a successful data project outlined below, your data team can uncover powerful insights that often go untapped.
Element #1: Deliberate Dialogue
Acknowledge that you might not have the right context for the data. And that’s true whether you are the data team or the product team. You need to commit to being open and listening actively.
It’s easy to quickly jump to conclusions and say, “We have analytics and insights.” However, moving too fast might mean failing to address biases that cause us to look at the data in a specific way.
Instead, we should carefully examine if any cognitive biases are impacting our insights and actively listen to our stakeholders.
Most of us have debates in our organizations; debates aren’t about listening. The goal of a debate is to get your point across no matter what. While someone else is talking, you might be thinking about how you’re going to contradict their point to prove your own.
So, deliberate dialogue versus debate is a big step in the right direction.
Element #2: Managing Uncertainty
We live in a volatile, complex, ambiguous and uncertain world. That means our brains are navigating uncertainty. You need to educate leaders that data always has an element of uncertainty.
We could run a model about the segmentation of our customers and whether or not they’ll buy our product. The output might indicate a 95% confidence interval, which isn’t certain. It’s best if all the decision-makers understand that everything is a probability.
There is no situation where we can say there is a 100% chance that investing in a specific demographic will increase our revenue.
You have to acknowledge that you will never have all the data, and you will never have all the answers. We don’t have any data on what will happen in the future. It’s not there.
So, we have to make predictions. We have to use foresight and strategy to guess what will happen.
And, if you’re data literate, you’re more likely to know if your strategy is a good bet.
Simply put, we manage uncertainty by seeing our efforts as reducing risk with data-driven insights.
Element #3: Knowing the Audience
It sounds basic. But suppose you’re an advanced analyst and do some complex modeling using advanced visualization (like a box plot).
But, your audience may tune out if you use the same level of visualization. They see numbers and lines. It’s about prioritizing the number and simplifying to show them that with a story.
Studies show people listen and attach emotionally to stories instead of just numbers. The first thing is to pick the proper visualization and pick the correct terminology to map to your stakeholder. You have to know their level. You have to know what’s in it for them.
Most importantly, you have to answer the question, “So what?”
If I’m sharing numbers from our latest marketing campaign to increase the marketing budget with the department responsible for paying the increased cost, the approach will be different from communicating with the creative team.
It comes down to knowing the stakeholder, not simplifying, but prioritizing what you’re going to show them. Everything else is just noise.
Element #4: Systems Thinking
Which numbers matter? That question is hard to answer because it’s not a simple one-to-one ratio. You can’t say, “This number drives this number,” and you’re done.
Twenty other numbers are impacting an outcome. If you’re lucky enough to have a data scientist, some algorithms help you understand which numbers have a higher correlation.
However, if you don’t, one of the non-technical elements of a successful data project you can employ is systems thinking.
The goal is to visualize a system map, which is the organization of all of those correlations that are related or not.
A personal example is weight management. What affects it? At the basic level, it’s calories and exercise, and you visualize that concept without any of the numbers.
If you communicate that well, you’ll have a dialogue with the stakeholder, and they’ll agree with you or disagree with you. The trick is you haven’t revealed any numbers yet; you’re just getting them to understand the system. Then, you can hit them with the numbers.
Here’s a simple example:
In a business context, let’s say sales revenue is down. So, I use active listening skills to uncover hypotheses from stakeholders as to why that might be the trend we’re seeing.
It could be because we’re discounting. It could be that we have fewer sales reps. It could be we’re charging more. It could be because the supply chain is slower. I could come up with 50 different scenarios that might impact sales. I’d write them all down.
Then, we analyze the data.
If we notice we’re selling more units, but we’re discounting 50% more, we can say, “We have the demand, we have the capacity, but we just need to change our discount strategy.”
I would focus on the discount number to share with my stakeholder.
The challenge is to have hundreds of variables and hundreds of hidden variables. It does become a little bit of an art and science. But using that simple model and slowly adding in complexity can help.
Element #5: A Thriving Data Culture
Another non-technical element of a successful data project is a company culture that supports data efforts. This environment is more likely to incorporate valuable insights rather than ignore them.
But how do you influence a data culture without being the CEO or in leadership?
Start by picking a data project that will not affect the company massively if it fails but is strategic enough to have a positive impact if it succeeds.
When you complete it, you’ll have a win report. You’ll share the information with the stakeholders and leaders to demonstrate value in work.
When possible, it helps to start data projects at the beginning of the pipeline, which is:
- Raw data
- Data lineage
- Data management
- Data governance
- Data at integration
We could build an excellent predictive model or create a compelling measurement framework with the right key performance indicators, but it’s a waste of time if we don’t trust the data; no one will use it.
Build that trust first with a quick win on the front end of the pipeline or the first part of the pipeline by connecting that data and trying to do some things around strategy.
Agile organizations making data-driven decisions are the ones that may fail, but they get back up. I like to call it to fall fast (because you can always get back up when you fall). The purpose is to learn and make incremental improvements.
If you want to learn how to gain insights you can act on and solve business problems with data—all while building a data-driven culture at your organization—sign up for Pragmatic Institute’s new course, Data Science for Business Leaders. Find out more at PragmaticInstitute.com/Data